SS-SFDA : Self-Supervised Source-Free Domain Adaptation for Road Segmentation in Hazardous Environments

Published in

(2020)

External Links:

Generate Graph DownloadPDF

TL;DR

This work presents a novel approach for unsupervised road segmentation in adverse weather conditions such as rain or fog using a new algorithm for source-free domain adaptation (SFDA) using self-supervised learning.

Abstract

We present a novel approach for unsupervised road segmentation in adverse weather conditions such as rain or fog. This includes a new algorithm for source-free domain adaptation (SFDA) using self-supervised learning. More-over, our approach uses several techniques to address various challenges in SFDA and improve performance, including online generation of pseudo-labels and self-attention as well as use of curriculum learning, entropy minimization and model distillation. We have evaluated the performance on 6 datasets corresponding to real and synthetic adverse weather conditions. Our method outperforms all prior works on unsupervised road segmentation and SFDA by at least 10.26%, and improves the training time by 18−180×. Moreover, our self-supervised algorithm exhibits similar accuracy performance in terms of mIOU score as compared to prior supervised methods.

Authors

Dinesh Manocha

8 papers

Rohan Chandra

3 papers

D. Kothandaraman

1 papers

References68 items

Bottleneck Transformers for Visual Recognition

Source-free Domain Adaptation via Distributional Alignment by Matching Batch Normalization Statistics

SoFA: Source-data-free Feature Alignment for Unsupervised Domain Adaptation

Domain Impression: A Source Data Free Domain Adaptation Method

Two-Phase Pseudo Label Densification for Self-training Based Domain Adaptation

SS-SFDA : Self-Supervised Source-Free Domain Adaptation for Road Segmentation in Hazardous Environments

Published in

(2020)

External Links:

Generate Graph DownloadPDF

TL;DR

Abstract

Authors

Dinesh Manocha

8 papers

Rohan Chandra

3 papers

D. Kothandaraman

1 papers

References68 items

Bottleneck Transformers for Visual Recognition

Source-free Domain Adaptation via Distributional Alignment by Matching Batch Normalization Statistics

SoFA: Source-data-free Feature Alignment for Unsupervised Domain Adaptation

Domain Impression: A Source Data Free Domain Adaptation Method

Two-Phase Pseudo Label Densification for Self-training Based Domain Adaptation

Unsupervised Domain Adaptation without Source Data by Casting a BAIT

BoMuDA: Boundless Multi-Source Domain Adaptive Segmentation in Unconstrained Environments

SNE-RoadSeg: Incorporating Surface Normal Information into Semantic Segmentation for Accurate Freespace Detection

Source Free Domain Adaptation with Image Translation

Domain Adaptation Without Source Data

Model Adaptation: Unsupervised Domain Adaptation Without Source Data

Source-Relaxed Domain Adaptation for Image Segmentation

Unsupervised Intra-Domain Adaptation for Semantic Segmentation Through Self-Supervision

Universal Source-Free Domain Adaptation

Do We Really Need to Access the Source Data? Source Hypothesis Transfer for Unsupervised Domain Adaptation

Curriculum Labeling: Revisiting Pseudo-Labeling for Semi-Supervised Learning

Generative Pseudo-label Refinement for Unsupervised Domain Adaptation

Forecasting Trajectory and Behavior of Road-Agents Using Spectral Clustering in Graph-LSTMs

Domain Bridge for Unpaired Image-to-Image Translation and Unsupervised Domain Adaptation

Reverse and Boundary Attention Network for Road Segmentation

Physics-Based Rendering for Improving Robustness to Rain

Pseudo-Labeling Curriculum for Unsupervised Domain Adaptation

RobustTP: End-to-End Trajectory Prediction for Heterogeneous Road-Agents in Dense Traffic with Noisy Sensor Inputs

Gated-SCNN: Gated Shape CNNs for Semantic Segmentation

RoadTrack: Realtime Tracking of Road Agents in Dense and Heterogeneous Environments

Learning to Self-Train for Semi-Supervised Few-Shot Classification

Semi-Supervised Domain Adaptation via Minimax Entropy

On The Power of Curriculum Learning in Training Deep Networks

DADA: Depth-Aware Domain Adaptation in Semantic Segmentation

Context-Reinforced Semantic Segmentation

Guided Curriculum Model Adaptation and Uncertainty-Aware Evaluation for Semantic Nighttime Image Segmentation

Curriculum Model Adaptation with Synthetic and Real Data for Semantic Foggy Scene Understanding

I Can See Clearly Now: Image Restoration via De-Raining

TraPHic: Trajectory Prediction in Dense and Heterogeneous Traffic Using Weighted Interactions

Source Free Domain Adaptation Using an Off-the-Shelf Classifier

ADVENT: Adversarial Entropy Minimization for Domain Adaptation in Semantic Segmentation

Dual Attention Network for Scene Segmentation

Unsupervised Domain Adaptation for Semantic Segmentation via Class-Balanced Self-training

Model Adaptation with Synthetic and Real Data for Semantic Dense Foggy Scene Understanding

Context Contrasted Feature and Gated Multi-scale Aggregation for Scene Segmentation

Self-Attention Generative Adversarial Networks

BDD100K: A Diverse Driving Video Database with Scalable Annotation Tooling

Context Encoding for Semantic Segmentation

Learning to Adapt Structured Output Space for Semantic Segmentation

Unsupervised Domain Adaptation for Semantic Segmentation with GANs

CyCADA: Cycle-Consistent Adversarial Domain Adaptation

Semantic Foggy Scene Understanding with Synthetic Data

Curriculum Domain Adaptation for Semantic Segmentation of Urban Scenes

The Raincouver Scene Parsing Benchmark for Self-Driving in Adverse Weather and at Night

Rethinking Atrous Convolution for Semantic Image Segmentation

Dilated Residual Networks

Embedding structured contour and location prior in siamesed fully convolutional networks for road detection

No More Discrimination: Cross City Adaptation of Road Scene Segmenters

FCNs in the Wild: Pixel-level Adversarial and Constraint-based Adaptation

Pyramid Scene Parsing Network

DeepLab: Semantic Image Segmentation with Deep Convolutional Nets, Atrous Convolution, and Fully Connected CRFs

The Cityscapes Dataset for Semantic Urban Scene Understanding

Fully convolutional networks for semantic segmentation

Driving in fog: the effects of driving experience and visibility on speed compensation and hazard avoidance.

Pedestrian Detection: An Evaluation of the State of the Art

Curriculum learning

Semi-supervised Learning by Entropy Minimization

Self-driving vehicle technology: progress and promises

International conference on machine learning

Superpixel-based Road Segmentation for Real-time Systems using CNN

Pseudo-Label : The Simple and Efficient Semi-Supervised Learning Method for Deep Neural Networks

Generative pseudo-label reﬁnement

Curriculum model adaptation with synthetic and 3050

Field of Study

Computer Science

Journal Information

Name

2021 IEEE/CVF International Conference on Computer Vision Workshops (ICCVW)

Page

3489-3497